full break down of population level effects

core demographics table

control PD prodromal p
(N=161) (N=505) (N=395)
subjectAge_BL 60.2 ± 11.6 62.3 ± 9.6 62.7 ± 7.1 0.012
subjectSex < 0.001
    F 59 (36.6%) 198 (39.2%) 215 (54.4%)
    M 102 (63.4%) 307 (60.8%) 180 (45.6%)
mds_updrs_part_i_summary_score 2.9 ± 3.1 6.6 ± 4.9 5.6 ± 4.5 < 0.001
mds_updrs_part_ii_summary_score 0.4 ± 1.0 6.4 ± 4.8 1.3 ± 2.5 < 0.001
mds_updrs_part_iii_summary_score 0.9 ± 1.9 20.8 ± 9.1 3.0 ± 4.2 < 0.001
code_upd2hy_hoehn_and_yahr_stage < 0.001
    0 160 (99.4%) 110 (21.8%) 378 (95.7%)
    1 1 ( 0.6%) 163 (32.3%) 13 ( 3.3%)
    2 0 ( 0.0%) 227 (45.0%) 4 ( 1.0%)
    3 0 ( 0.0%) 5 ( 1.0%) 0 ( 0.0%)
geneticMutation < 0.001
    Absent 161 (100.0%) 349 (69.1%) 62 (15.7%)
    GBA 0 ( 0.0%) 44 ( 8.7%) 166 (42.0%)
    LRRK2 0 ( 0.0%) 112 (22.2%) 167 (42.3%)
has24 < 0.001
    FALSE 155 (96.3%) 398 (78.8%) 372 (94.2%)
    TRUE 6 ( 3.7%) 107 (21.2%) 23 ( 5.8%)
has48 < 0.001
    FALSE 158 (98.1%) 469 (92.9%) 324 (82.0%)
    TRUE 3 ( 1.9%) 36 ( 7.1%) 71 (18.0%)

sex, genetics, model, DX

reproducibility: ICC

reproducibility: Mean absolute difference

reproducibility: Correlation vs Model-adjusted correlation

raw corr

mod-adj corr

discrimination: control vs prodromal vs PD

best model: confusion matrix - cross-pred - raw accuracy - 3-class

best model: confusion matrix - cross-pred - raw accuracy - 2-class

summarize overall improvement by class with measurement type / dx regression

effects of mfg.model

longitudinal scanner effect

longitudinal connectivity study

longitudinal roi study

roi study results

insula connectivity study results